some software optimizations

from:
https://git.do.srb2.org/Hanicef/SRB2Classic/-/merge_requests/6#d4bba40be8388ccd557d03ef4e4a36d0728c7751_645_587

minus the buffer flipping change, karts resolution stuff is hell and even worse due to birdmod things
maybe will look into that sometime
This commit is contained in:
Alug committed 2025-04-07 08:35:33 +02:00
1 parent 7cf1e8f15f
commit 9902cac3ee
3 files changed
+68 -75

No files matched your search

+61 -68
View File
@@ -25,9 +25,9 @@
void R_DrawColumn_8(void)
{
INT32 count;
register UINT8 *dest;
register fixed_t frac;
fixed_t fracstep;
UINT8 *restrict dest;
intptr_t frac;
intptr_t fracstep;
INT32 npow2min;
INT32 npow2max;
@@ -57,10 +57,10 @@ void R_DrawColumn_8(void)
// Inner loop that does the actual texture mapping, e.g. a DDA-like scaling.
// This is as fast as it gets.
register const UINT8 *source = dc_source;
register const lighttable_t *colormap = dc_colormap;
const UINT8 *restrict source = dc_source;
const lighttable_t *restrict colormap = dc_colormap;
register INT32 heightmask = dc_sourcelength-1;
intptr_t heightmask = dc_sourcelength-1;
npow2min = -1;
npow2max = dc_sourcelength;
@@ -77,10 +77,7 @@ void R_DrawColumn_8(void)
}
else
{
while (frac >= heightmask)
{
frac -= heightmask;
}
frac %= heightmask;
}
do
@@ -105,20 +102,20 @@ void R_DrawColumn_8(void)
dest += vid.width;
#if __SIZEOF_POINTER__ < 8 // 64-bit systems have large enough numbers for this to be a non-issue
// Avoid overflow.
if (fracstep > 0x7FFFFFFF - frac)
{
frac += fracstep - heightmask;
}
else
#endif
{
frac += fracstep;
}
while (frac >= heightmask)
{
frac -= heightmask;
}
frac %= heightmask;
} while (--count);
}
else
@@ -147,9 +144,10 @@ void R_DrawColumn_8(void)
void R_Draw2sMultiPatchColumn_8(void)
{
INT32 count;
register UINT8 *dest;
register fixed_t frac;
fixed_t fracstep;
UINT8 *restrict dest;
intptr_t frac;
intptr_t fracstep;
INT32 npow2min;
INT32 npow2max;
@@ -178,10 +176,11 @@ void R_Draw2sMultiPatchColumn_8(void)
// Inner loop that does the actual texture mapping, e.g. a DDA-like scaling.
// This is as fast as it gets.
register const UINT8 *source = dc_source;
register const lighttable_t *colormap = dc_colormap;
register INT32 heightmask = dc_sourcelength-1;
register UINT8 val;
const UINT8 *restrict source = dc_source;
const lighttable_t *restrict colormap = dc_colormap;
intptr_t heightmask = dc_sourcelength-1;
UINT8 val;
npow2min = -1;
npow2max = dc_sourcelength;
@@ -198,10 +197,7 @@ void R_Draw2sMultiPatchColumn_8(void)
}
else
{
while (frac >= heightmask)
{
frac -= heightmask;
}
frac %= heightmask;
}
do
@@ -233,19 +229,18 @@ void R_Draw2sMultiPatchColumn_8(void)
dest += vid.width;
// Avoid overflow.
#if __SIZEOF_POINTER__ < 8
if (fracstep > 0x7FFFFFFF - frac)
{
frac += fracstep - heightmask;
}
else
#endif
{
frac += fracstep;
}
while (frac >= heightmask)
{
frac -= heightmask;
}
frac %= heightmask;
} while (--count);
}
else
@@ -285,9 +280,10 @@ void R_Draw2sMultiPatchColumn_8(void)
void R_Draw2sMultiPatchTranslucentColumn_8(void)
{
INT32 count;
register UINT8 *dest;
register fixed_t frac;
fixed_t fracstep;
UINT8 *restrict dest;
intptr_t frac;
intptr_t fracstep;
INT32 npow2min;
INT32 npow2max;
@@ -316,11 +312,12 @@ void R_Draw2sMultiPatchTranslucentColumn_8(void)
// Inner loop that does the actual texture mapping, e.g. a DDA-like scaling.
// This is as fast as it gets.
register const UINT8 *source = dc_source;
register const UINT8 *transmap = dc_transmap;
register const lighttable_t *colormap = dc_colormap;
register INT32 heightmask = dc_sourcelength-1;
const UINT8 *restrict source = dc_source;
const UINT8 *transmap = dc_transmap;
const lighttable_t *restrict colormap = dc_colormap;
intptr_t heightmask = dc_sourcelength-1;
register UINT8 val;
npow2min = -1;
npow2max = dc_sourcelength;
@@ -337,10 +334,7 @@ void R_Draw2sMultiPatchTranslucentColumn_8(void)
}
else
{
while (frac >= heightmask)
{
frac -= heightmask;
}
frac %= heightmask;
}
do
@@ -372,19 +366,18 @@ void R_Draw2sMultiPatchTranslucentColumn_8(void)
dest += vid.width;
// Avoid overflow.
#if __SIZEOF_POINTER__ < 8
if (fracstep > 0x7FFFFFFF - frac)
{
frac += fracstep - heightmask;
}
else
#endif
{
frac += fracstep;
}
while (frac >= heightmask)
{
frac -= heightmask;
}
frac %= heightmask;
} while (--count);
}
else
@@ -470,8 +463,10 @@ void R_DrawShadeColumn_8(void)
void R_DrawTranslucentColumn_8(void)
{
register INT32 count;
register UINT8 *dest;
register fixed_t frac, fracstep;
UINT8 *restrict dest;
intptr_t frac;
intptr_t fracstep;
INT32 npow2min;
INT32 npow2max;
@@ -496,10 +491,11 @@ void R_DrawTranslucentColumn_8(void)
// Inner loop that does the actual texture mapping, e.g. a DDA-like scaling.
// This is as fast as it gets.
register const UINT8 *source = dc_source;
register const UINT8 *transmap = dc_transmap;
register const lighttable_t *colormap = dc_colormap;
register INT32 heightmask = dc_sourcelength-1;
const UINT8 *restrict source = dc_source;
const UINT8 *transmap = dc_transmap;
const lighttable_t *restrict colormap = dc_colormap;
intptr_t heightmask = dc_sourcelength-1;
npow2min = -1;
npow2max = dc_sourcelength;
@@ -516,10 +512,7 @@ void R_DrawTranslucentColumn_8(void)
}
else
{
while (frac >= heightmask)
{
frac -= heightmask;
}
frac %= heightmask;
}
do
@@ -579,8 +572,10 @@ void R_DrawTranslucentColumn_8(void)
void R_DrawTranslatedTranslucentColumn_8(void)
{
register INT32 count;
register UINT8 *dest;
register fixed_t frac, fracstep;
UINT8 *restrict dest;
intptr_t frac;
intptr_t fracstep;
INT32 npow2min;
INT32 npow2max;
@@ -605,7 +600,8 @@ void R_DrawTranslatedTranslucentColumn_8(void)
// Inner loop that does the actual texture mapping, e.g. a DDA-like scaling.
// This is as fast as it gets.
register INT32 heightmask = dc_sourcelength-1;
intptr_t heightmask = dc_sourcelength-1;
npow2min = -1;
npow2max = dc_sourcelength;
@@ -622,10 +618,7 @@ void R_DrawTranslatedTranslucentColumn_8(void)
}
else
{
while (frac >= heightmask)
{
frac -= heightmask;
}
frac %= heightmask;
}
do
@@ -739,15 +732,15 @@ void R_DrawTranslatedColumn_8(void)
*/
void R_DrawSpan_8 (void)
{
fixed_t xposition;
fixed_t yposition;
fixed_t xstep, ystep;
uintptr_t xposition;
uintptr_t yposition;
uintptr_t xstep, ystep;
register UINT32 bit;
UINT8 *source;
UINT8 *colormap;
register UINT8 *dest;
const UINT8 *deststop = screens[0] + vid.rowbytes * vid.height;
UINT8 *restrict source;
UINT8 *restrict colormap;
UINT8 *restrict dest;
const UINT8 *restrict deststop = screens[0] + vid.rowbytes * vid.height;
register size_t count = (ds_x2 - ds_x1 + 1);
size_t i;
+6 -6
View File
@@ -322,7 +322,7 @@ void R_RenderMaskedSegRange(drawseg_t *ds, INT32 x1, INT32 x2)
if (!ldef->alpha)
return;
if (ldef->blendmode)
{
if (ldef->alpha == NUMTRANSMAPS || ldef->blendmode == AST_MODULATE)
@@ -628,7 +628,7 @@ void R_RenderMaskedSegRange(drawseg_t *ds, INT32 x1, INT32 x2)
// draw the texture
col = (column_t *)((UINT8 *)R_GetColumn(texnum, maskedtexturecol[dc_x]) - 3);
colfunc_2s(col);
}
spryscale += rw_scalestep;
@@ -1581,7 +1581,7 @@ static void R_RenderSegLoop (void)
{
if (curline->polyseg && (ffloor[i].polyobj != curline->polyseg))
continue;
ffloor[i].f_clip[rw_x] = ffloor[i].c_clip[rw_x] = (INT16)((ffloor[i].b_frac >> HEIGHTBITS) & 0xFFFF);
ffloor[i].b_frac += ffloor[i].b_step;
}
@@ -1590,7 +1590,7 @@ static void R_RenderSegLoop (void)
topfrac += topstep;
bottomfrac += bottomstep;
}
//colfunc = wallcolfunc;
}
@@ -1690,7 +1690,7 @@ void R_StoreWallRange(INT32 start, INT32 stop)
while (need > maxopenings);
openings = Z_Realloc(openings, maxopenings * sizeof (*openings), PU_STATIC, NULL);
lastopening = openings + pos;
if (oldopenings == NULL)
return;
@@ -2817,7 +2817,7 @@ void R_StoreWallRange(INT32 start, INT32 stop)
R_ExpandPlane(ffloor[i].plane, rw_x, rw_stopx - 1);
}
}
rw_silhouette = &(ds_p->silhouette);
rw_tsilheight = &(ds_p->tsilheight);
rw_bsilheight = &(ds_p->bsilheight);
+1 -1
View File
@@ -897,7 +897,7 @@ static void R_DrawVisSprite(vissprite_t *vis)
else
{
// Non-paper drawing loop
for (dc_x = vis->x1; dc_x <= vis->x2; dc_x++, frac += vis->xiscale)
for (dc_x = vis->x1; dc_x <= vis->x2 && (frac>>FRACBITS) < patch->width; dc_x++, frac += vis->xiscale)
{
texturecolumn = CLAMP(frac >> FRACBITS, 0, SHORT(patch->width) - 1);
column = (column_t *)((UINT8 *)patch + LONG(patch->columnofs[texturecolumn]));